From e17117310b73ce6d2340ad46a539d3896a2d6de8 Mon Sep 17 00:00:00 2001 From: Philipp Reisner Date: Mon, 27 Jun 2011 11:51:46 +0200 Subject: [PATCH] drbd: Replaced md_io_mutex by an atomic: md_io_in_use The new function drbd_md_get_buffer() aborts waiting for the buffer in case the disk failes in the meantime. Signed-off-by: Philipp Reisner Signed-off-by: Lars Ellenberg --- drivers/block/drbd/drbd_actlog.c | 44 +++++++++++++++++++++++++++++++--------- drivers/block/drbd/drbd_int.h | 6 ++++-- drivers/block/drbd/drbd_main.c | 19 ++++++++++------- drivers/block/drbd/drbd_worker.c | 1 + 4 files changed, 51 insertions(+), 19 deletions(-) diff --git a/drivers/block/drbd/drbd_actlog.c b/drivers/block/drbd/drbd_actlog.c index c64ca6b..4271352 100644 --- a/drivers/block/drbd/drbd_actlog.c +++ b/drivers/block/drbd/drbd_actlog.c @@ -65,6 +65,23 @@ struct drbd_atodb_wait { int w_al_write_transaction(struct drbd_conf *, struct drbd_work *, int); +void *drbd_md_get_buffer(struct drbd_conf *mdev) +{ + int r; + + wait_event(mdev->misc_wait, + (r = atomic_cmpxchg(&mdev->md_io_in_use, 0, 1)) == 0 || + mdev->state.disk <= D_FAILED); + + return r ? NULL : page_address(mdev->md_io_page); +} + +void drbd_md_put_buffer(struct drbd_conf *mdev) +{ + if (atomic_dec_and_test(&mdev->md_io_in_use)) + wake_up(&mdev->misc_wait); +} + static int _drbd_md_sync_page_io(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, struct page *page, sector_t sector, @@ -90,6 +107,7 @@ static int _drbd_md_sync_page_io(struct drbd_conf *mdev, bio->bi_end_io = drbd_md_io_complete; bio->bi_rw = rw; + atomic_inc(&mdev->md_io_in_use); /* drbd_md_put_buffer() is in the completion handler */ if (drbd_insert_fault(mdev, (rw & WRITE) ? DRBD_FAULT_MD_WR : DRBD_FAULT_MD_RD)) bio_endio(bio, -EIO); else @@ -109,7 +127,7 @@ int drbd_md_sync_page_io(struct drbd_conf *mdev, struct drbd_backing_dev *bdev, int offset = 0; struct page *iop = mdev->md_io_page; - D_ASSERT(mutex_is_locked(&mdev->md_io_mutex)); + D_ASSERT(atomic_read(&mdev->md_io_in_use) == 1); BUG_ON(!bdev->md_bdev); @@ -326,8 +344,13 @@ w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) return 1; } - mutex_lock(&mdev->md_io_mutex); /* protects md_io_buffer, al_tr_cycle, ... */ - buffer = (struct al_transaction *)page_address(mdev->md_io_page); + buffer = drbd_md_get_buffer(mdev); /* protects md_io_buffer, al_tr_cycle, ... */ + if (!buffer) { + dev_err(DEV, "disk failed while waiting for md_io buffer\n"); + complete(&((struct update_al_work *)w)->event); + put_ldev(mdev); + return 1; + } buffer->magic = __constant_cpu_to_be32(DRBD_MAGIC); buffer->tr_number = cpu_to_be32(mdev->al_tr_number); @@ -372,7 +395,7 @@ w_al_write_transaction(struct drbd_conf *mdev, struct drbd_work *w, int unused) D_ASSERT(mdev->al_tr_pos < MD_AL_MAX_SIZE); mdev->al_tr_number++; - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); complete(&((struct update_al_work *)w)->event); put_ldev(mdev); @@ -441,8 +464,9 @@ int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) /* lock out all other meta data io for now, * and make sure the page is mapped. */ - mutex_lock(&mdev->md_io_mutex); - buffer = page_address(mdev->md_io_page); + buffer = drbd_md_get_buffer(mdev); + if (!buffer) + return 0; /* Find the valid transaction in the log */ for (i = 0; i <= mx; i++) { @@ -450,7 +474,7 @@ int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) if (rv == 0) continue; if (rv == -1) { - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); return 0; } cnr = be32_to_cpu(buffer->tr_number); @@ -476,7 +500,7 @@ int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) if (!found_valid) { dev_warn(DEV, "No usable activity log found.\n"); - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); return 1; } @@ -491,7 +515,7 @@ int drbd_al_read_log(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) rv = drbd_al_read_tr(mdev, bdev, buffer, i); ERR_IF(rv == 0) goto cancel; if (rv == -1) { - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); return 0; } @@ -532,7 +556,7 @@ cancel: mdev->al_tr_pos = 0; /* ok, we are done with it */ - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); dev_info(DEV, "Found %d transactions (%d active extents) in activity log.\n", transactions, active_extents); diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h index 91e69ff..55cae74 100644 --- a/drivers/block/drbd/drbd_int.h +++ b/drivers/block/drbd/drbd_int.h @@ -1095,7 +1095,7 @@ struct drbd_conf { struct page *md_io_page; /* one page buffer for md_io */ struct page *md_io_tmpp; /* for logical_block_size != 512 */ struct drbd_md_io md_io; - struct mutex md_io_mutex; /* protects the md_io, md_io_page and md_io_tmpp */ + atomic_t md_io_in_use; /* protects the md_io, md_io_page and md_io_tmpp */ spinlock_t al_lock; wait_queue_head_t al_wait; struct lru_cache *act_log; /* activity log */ @@ -1537,8 +1537,10 @@ extern void resume_next_sg(struct drbd_conf *mdev); extern void suspend_other_sg(struct drbd_conf *mdev); extern int drbd_resync_finished(struct drbd_conf *mdev); /* maybe rather drbd_main.c ? */ +extern void *drbd_md_get_buffer(struct drbd_conf *mdev); +extern void drbd_md_put_buffer(struct drbd_conf *mdev); extern int drbd_md_sync_page_io(struct drbd_conf *mdev, - struct drbd_backing_dev *bdev, sector_t sector, int rw); + struct drbd_backing_dev *bdev, sector_t sector, int rw); extern void drbd_ov_oos_found(struct drbd_conf*, sector_t, int); extern void drbd_rs_controller_reset(struct drbd_conf *mdev); diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 16969c2..1b59ab3 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -3043,8 +3043,8 @@ void drbd_init_set_defaults(struct drbd_conf *mdev) atomic_set(&mdev->rs_sect_in, 0); atomic_set(&mdev->rs_sect_ev, 0); atomic_set(&mdev->ap_in_flight, 0); + atomic_set(&mdev->md_io_in_use, 0); - mutex_init(&mdev->md_io_mutex); mutex_init(&mdev->data.mutex); mutex_init(&mdev->meta.mutex); sema_init(&mdev->data.work.s, 0); @@ -3722,8 +3722,10 @@ void drbd_md_sync(struct drbd_conf *mdev) if (!get_ldev_if_state(mdev, D_FAILED)) return; - mutex_lock(&mdev->md_io_mutex); - buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page); + buffer = drbd_md_get_buffer(mdev); + if (!buffer) + goto out; + memset(buffer, 0, 512); buffer->la_size = cpu_to_be64(drbd_get_capacity(mdev->this_bdev)); @@ -3754,7 +3756,8 @@ void drbd_md_sync(struct drbd_conf *mdev) * since we updated it on metadata. */ mdev->ldev->md.la_size_sect = drbd_get_capacity(mdev->this_bdev); - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); +out: put_ldev(mdev); } @@ -3774,8 +3777,9 @@ int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) if (!get_ldev_if_state(mdev, D_ATTACHING)) return ERR_IO_MD_DISK; - mutex_lock(&mdev->md_io_mutex); - buffer = (struct meta_data_on_disk *)page_address(mdev->md_io_page); + buffer = drbd_md_get_buffer(mdev); + if (!buffer) + goto out; if (!drbd_md_sync_page_io(mdev, bdev, bdev->md.md_offset, READ)) { /* NOTE: can't do normal error processing here as this is @@ -3836,7 +3840,8 @@ int drbd_md_read(struct drbd_conf *mdev, struct drbd_backing_dev *bdev) mdev->sync_conf.al_extents = 127; err: - mutex_unlock(&mdev->md_io_mutex); + drbd_md_put_buffer(mdev); + out: put_ldev(mdev); return rv; diff --git a/drivers/block/drbd/drbd_worker.c b/drivers/block/drbd/drbd_worker.c index eedfe31..94a128b 100644 --- a/drivers/block/drbd/drbd_worker.c +++ b/drivers/block/drbd/drbd_worker.c @@ -78,6 +78,7 @@ void drbd_md_io_complete(struct bio *bio, int error) md_io->error = error; complete(&md_io->event); + drbd_md_put_buffer(mdev); } /* reads on behalf of the partner, -- 2.7.4